forecasting

Creates, updates, deletes, gets or lists a forecasting resource.

Overview

Name	forecasting
Type	Resource
Id	databricks_workspace.ml.forecasting

Fields

The following fields are returned by SELECT queries:

get

Name	Datatype	Description
experiment_id	string	The unique ID for the forecasting experiment.
experiment_page_url	string	The URL to the forecasting experiment page.
state	string	The current state of the forecasting experiment. (CANCELLED, FAILED, PENDING, RUNNING, SUCCEEDED)

Methods

The following methods are available for this resource:

Name	Accessible by	Required Params	Optional Params	Description
get	select	experiment_id, deployment_name		Public RPC to get forecasting experiment
create	insert	deployment_name, train_data_path, target_column, time_column, forecast_granularity, forecast_horizon		Creates a serverless forecasting experiment. Returns the experiment ID.

Parameters

Parameters can be passed in the WHERE clause of a query. Check the Methods section to see which parameters are required or optional for each operation.

Name	Datatype	Description
deployment_name	string	The Databricks Workspace Deployment Name (default: dbc-abcd0123-a1bc)
experiment_id	string	The unique ID of a forecasting experiment

SELECT examples

get

Public RPC to get forecasting experiment

SELECT
experiment_id,
experiment_page_url,
state
FROM databricks_workspace.ml.forecasting
WHERE experiment_id = '{{ experiment_id }}' -- required
AND deployment_name = '{{ deployment_name }}' -- required
;

INSERT examples

create

Creates a serverless forecasting experiment. Returns the experiment ID.

INSERT INTO databricks_workspace.ml.forecasting (
train_data_path,
target_column,
time_column,
forecast_granularity,
forecast_horizon,
custom_weights_column,
experiment_path,
future_feature_data_path,
holiday_regions,
include_features,
max_runtime,
prediction_data_path,
primary_metric,
register_to,
split_column,
timeseries_identifier_columns,
training_frameworks,
deployment_name
)
SELECT 
'{{ train_data_path }}' /* required */,
'{{ target_column }}' /* required */,
'{{ time_column }}' /* required */,
'{{ forecast_granularity }}' /* required */,
{{ forecast_horizon }} /* required */,
'{{ custom_weights_column }}',
'{{ experiment_path }}',
'{{ future_feature_data_path }}',
'{{ holiday_regions }}',
'{{ include_features }}',
{{ max_runtime }},
'{{ prediction_data_path }}',
'{{ primary_metric }}',
'{{ register_to }}',
'{{ split_column }}',
'{{ timeseries_identifier_columns }}',
'{{ training_frameworks }}',
'{{ deployment_name }}'
RETURNING
experiment_id,
experiment_page_url,
state
;

Manifest

# Description fields are for documentation purposes
- name: forecasting
props:
  - name: deployment_name
    value: "{{ deployment_name }}"
    description: Required parameter for the forecasting resource.
  - name: train_data_path
    value: "{{ train_data_path }}"
    description: |
      The fully qualified path of a Unity Catalog table, formatted as catalog_name.schema_name.table_name, used as training data for the forecasting model.
  - name: target_column
    value: "{{ target_column }}"
    description: |
      The column in the input training table used as the prediction target for model training. The values in this column are used as the ground truth for model training.
  - name: time_column
    value: "{{ time_column }}"
    description: |
      The column in the input training table that represents each row's timestamp.
  - name: forecast_granularity
    value: "{{ forecast_granularity }}"
    description: |
      The time interval between consecutive rows in the time series data. Possible values include: '1 second', '1 minute', '5 minutes', '10 minutes', '15 minutes', '30 minutes', 'Hourly', 'Daily', 'Weekly', 'Monthly', 'Quarterly', 'Yearly'.
  - name: forecast_horizon
    value: {{ forecast_horizon }}
    description: |
      The number of time steps into the future to make predictions, calculated as a multiple of forecast_granularity. This value represents how far ahead the model should forecast.
  - name: custom_weights_column
    value: "{{ custom_weights_column }}"
    description: |
      The column in the training table used to customize weights for each time series.
  - name: experiment_path
    value: "{{ experiment_path }}"
    description: |
      The path in the workspace to store the created experiment.
  - name: future_feature_data_path
    value: "{{ future_feature_data_path }}"
    description: |
      The fully qualified path of a Unity Catalog table, formatted as catalog_name.schema_name.table_name, used to store future feature data for predictions.
  - name: holiday_regions
    value:
      - "{{ holiday_regions }}"
    description: |
      The region code(s) to automatically add holiday features. Currently supports only one region.
  - name: include_features
    value:
      - "{{ include_features }}"
    description: |
      Specifies the list of feature columns to include in model training. These columns must exist in the training data and be of type string, numerical, or boolean. If not specified, no additional features will be included. Note: Certain columns are automatically handled: - Automatically excluded: split_column, target_column, custom_weights_column. - Automatically included: time_column.
  - name: max_runtime
    value: {{ max_runtime }}
    description: |
      The maximum duration for the experiment in minutes. The experiment stops automatically if it exceeds this limit.
  - name: prediction_data_path
    value: "{{ prediction_data_path }}"
    description: |
      The fully qualified path of a Unity Catalog table, formatted as catalog_name.schema_name.table_name, used to store predictions.
  - name: primary_metric
    value: "{{ primary_metric }}"
    description: |
      The evaluation metric used to optimize the forecasting model.
  - name: register_to
    value: "{{ register_to }}"
    description: |
      The fully qualified path of a Unity Catalog model, formatted as catalog_name.schema_name.model_name, used to store the best model.
  - name: split_column
    value: "{{ split_column }}"
    description: |
      // The column in the training table used for custom data splits. Values must be 'train', 'validate', or 'test'.
  - name: timeseries_identifier_columns
    value:
      - "{{ timeseries_identifier_columns }}"
    description: |
      The column in the training table used to group the dataset for predicting individual time series.
  - name: training_frameworks
    value:
      - "{{ training_frameworks }}"
    description: |
      List of frameworks to include for model tuning. Possible values are 'Prophet', 'ARIMA', 'DeepAR'. An empty list includes all supported frameworks.